ZestFinance Inc., known as Zest AI, is boosting its machine-learning credit underwriting capabilities with FairBoost, the latest edition to the company’s suite of tools aimed at minimizing risk and maximizing fairness for lenders.
The Burbank-based fintech company, which develops artificial intelligence-driven lending infrastructures, tested the new tool with two large companies and intends to expand access to financial institutions on a larger scale in coming quarters. The company sells its technology to primarily credit union customers.
“We have been purposeful about building AI-driven models of all sizes, but that are making a difference,” Zest AI Chief Executive Mike de Vere said. “Because what we believe is every American deserves a fair shot to affordable credit.”
Zest AI, which was founded in 2009, automates the underwriting process for financial institutions. The company’s goal is to expand credit access by making the lending process fairer and more transparent.
When applying for a loan in the U.S., the credit scoring system determines how likely a person is to pay his or her bills on time. The FICO score is the most widely used model by lenders, developed to eliminate bias by relying solely on credit history, not identifying data. But critics say credit bureaus’ reliance on limited historical financial data unfairly amplifies past inequalities to loan access.
Federal Reserve data shows more than one in five Black Americans and one in nine Hispanic Americans have FICO scores below 620 — a level most lenders considers to be high risk. Only one in 19 White Americans have the same score level.
To combat credit disparity from generational wealth gaps along the lines of race, sex and national origin, the Consumer Financial Protection Bureau requires lenders to pass a fair-lending compliance test. This requires entities to search and try all applicable credit formulas before determining a loan application’s risk.
FairBoost acts as the search engine for these less-discriminatory alternative models. According to de Vere, the tool consumes hundreds of variables about a borrower then runs all possible models to make a thorough, fair assessment.
As the world of consumer credit underwriting automates, Zest AI says this machine learning model cracked the code for minimizing risk while maximizing fairness.
“We have the patent for debiasing a model,” de Vere said.
The company has more than 370 AI models in production, and is working with large financial firms including Citibank N.A. and Truist Financial Corp., as well as local credit unions.
Zest’s AI-automated underwriting insights caught the eye of Sync1 Systems, LLC, which operates a cloud-based loan origination system. The financial technology company, based in Austin, Texas, recently announced that Zest AI would be available on its platforms.
“We are excited to partner with Zest AI to offer our clients a more advanced and effective lending platform,” said Steve Maloney, the chief executive of Sync1 Systems.
“We are always looking for ways to improve our technology and provide our clients with the best possible lending experience. This partnership with Zest AI is a significant step forward in achieving that goal,” he added.
By using more ingredients in its prediction system, Zest AI’s models allow its banking customers to holistically explain loan decisions while updating with economic fluctuations. According to de Vere, the inclusivity of credit compounds with every new model the company builds.
The company says lenders see an average 25% increase in approvals without increasing their risk tolerance when using Zest AI’s algorithms. The company emphasizes the technology works to close the access gap, not to raise approvals across the board.